---
title: "Nanjing University's TempR1 Teaches One Video AI to Master All Five Timestamp Tasks"
subtitle: "A new multi-task reinforcement learning recipe teaches video language models to pinpoint events in time across five different temporal tasks at once."
url: "https://alphasignal.ai/news/nanjing-university-s-tempr1-teaches-one-video-ai-to-master-all-five-timestamp"
author: "AlphaSignal Newsroom"
author_url: "https://alphasignal.ai/editorial-team"
company: "AlphaSignal"
company_url: "https://alphasignal.ai/company/alphasignal"
published: "2026-04-14T00:00:00.000Z"
updated: "2026-08-23T22:04:57.700Z"
type: "Paper"
topics:
  - "Video"
  - "Post Training"
subtopics:
  - "Video Understanding"
  - "Rl"
upvotes: 1832
source_url: "https://arxiv.org/abs/2512.03963"
comments: 0
truncated: true
---

# Nanjing University's TempR1 Teaches One Video AI to Master All Five Timestamp Tasks

> A new multi-task reinforcement learning recipe teaches video language models to pinpoint events in time across five different temporal tasks at once.

Video language models are getting fluent at describing what happens on screen, but ask them exactly when something happens and things fall apart quickly. A new framework called [TempR1](https://arxiv.org/abs/2512.03963), from researchers at Nanjing University, ByteDance, and Shanghai AI Lab, trains a single MLLM to handle five time-related video tasks simultaneously using reinforcement learning with rewards tailored to each task's temporal structure.

Built on Qwen2.5-VL-7B and trained on roughly 60K curated samples spanning temporal grounding, dense temporal grounding, temporal action localization, video highlight detection, and grounded video QA, TempR1 posts state-of-the-art results across the board, including a 13-point jump in mF1 over MUSEG on ActivityNet-v1.3 action localization.

## Why timestamps break MLLMs

Most current video MLLMs are trained on general video-text alignment and struggle to output precise time intervals. A handful of recent reinforcement-learning approaches have improved this, but they mostly focus on static images or short clips, with limited exploration of long-form temporal reasoning beyond basic grounding. That single-task tunnel vision produces models that can localize one described event but choke when asked to enumerate every instance of an action or provide temporal evidence for a QA answer.

Supervised fine-tuning has its own trap. Rigid token-level supervision on temporal instruction data tends to overfit and erodes the base model's general reasoning ability. RL-based fine-tuning, which rewards any temporally and semantically plausible prediction rather than one exact string, has emerged as the better path, though the community has been building one narrow reward per task.

_Preview only. Full text requires an AlphaSignal Pro account._

## Sources

- [TempR1: Improving Temporal Understanding of MLLMs via Time-Aware Multi-Task Reinforcement Learning](https://arxiv.org/html/2512.03963v3)
- [TempR1 CVPR 2026 paper](https://openaccess.thecvf.com/content/CVPR2026/papers/Wu_TempR1_Improving_Temporal_Understanding_of_MLLMs_via_Temporal-Aware_Multi-Task_Reinforcement_CVPR_2026_paper.pdf)
- [Qwen2.5-VL Technical Report](https://arxiv.org/abs/2502.13923)
- [TempSamp-R1: Effective Temporal Sampling with Reinforcement Fine-Tuning](https://arxiv.org/abs/2509.18056)